Research Methods in Finance II
Individual Project
Due: 2pm April 25
th
, 2025
Submission via LEARN AND Turnitin
Absolute max of words: 1500 +/- 10%
For this assignment you are provided with a large dataset covering publicly listed firms in the UK for
the period 2001 – 2023. The data files provided to you include (but you do not have to use all of them):
uk_isins.dta (a cross-sectional dataset of all UK listed firms),
uk_financials.dta (a panel dataset of financial statement variables for UK listed firms over 23 years),
uk_directors.dta (a cross-sectional dataset of individual characteristics of the directors who have served
on the boards of UK listed firms during the period),
uk_edu.dta (a cross-sectional dataset of directors’ education)
uk_directors_employment.dta (a cross-sectional dataset of employment history of directors in UK listed
firms),
uk_committees (a panel dataset of board committee composition of UK listed firms over 23 years),
uk_comp.dta (a panel dataset of board members compensation of UK listed firms over 23 years).
The original data sources are Capital IQ and BoardEx. Each student is assigned a subset of firms from
three industries based on the firm industry classification variable named ind_rmf2 in uk_isins.dta as
specified in the file rmf2_assigned_ind_2025.xlsx.
The goal is to perform statistical analysis of the relationship between an aspect of board diversity and
firm performance as measured by Tobin’s q. A recent review article of the existing literature examining
this topic is Knyazeva et al. (2021). Students are expected to study the article and find further reference
articles examined therein to substantiate their empirical design choices. A good example article as an
empirical design guideline is Bernile et al. (2018).
Marking Criteria:
Students are required to use the statistical software package covered in the course, i.e., Stata. Projects
using methods not covered in the course will not be given marks. Answers exceeding the maximum
word count provided after each question will only be marked up to the respective word limit. Variables
must be reported and referred to with consistent names and labels throughout the project. Tables and
figures need to be self-explanatory (readers must understand these just by looking at the table/graph
and referring to the associated notes beneath, without needing to refer to the body of the text).
This is an individual assignment. Similarities between projects will be severely penalized. Originality
and creativity will be rewarded. Each student has a unique subset of data assigned to them and therefore
their statistical results will differ. The discussion questions refer to the unique numerical results of each
student and the answers should point explicitly to the corresponding analysis, numbers and tables.
Full marks will be awarded to students who successfully address each question, adhere strictly to all
provided instructions and demonstrate originality. Exceptional responses may not need to use the entire
word limit. The quality of each answer will be the primary criterion for determining the marks awarded.
1
2
General Instructions:
When carrying out the project clearly write down the number of the question/task you are
answering/performing and order your answers sequentially. If numbers are not provided for a question,
zero marks will be allocated to that question.
The baseline regression model you will employ is as follows:
= 0 + 0 0 + 0 + 1 0 +∑ +∑ +
Where is the dependent variable Tobin’s q;
0 is a continuous explanatory variable of interest capturing an aspect of diversity;
is an indicator variable denoting the three industries assigned to you;
are control variables appropriate for the regression equation to be decided by you based on your
literature review and
are year dummies.
To choose your variable of interest follow the following steps:
i) Study the review article by Knyazeva et al. (2021)
ii) Select between one and three other articles cited therein to be your guiding references for this
project.
1
iii) Knyazeva et al. (2021) discusses several dimensions of diversity, including demographic
factors like gender, ethnicity, and cultural background, alongside other personal attributes such
as political beliefs and family or military background. There is also an emphasis on professional
diversity, which considers differences in industry experience and educational background. You
can also explore dimensions not covered in the article, for example, director nationality
provided to you in uk_directors.
Bonus credit of up to 20 marks will be awarded for creative diversity measures and/or ones that
require extra computational steps. You do NOT have to collect additional data for the project,
but NO penalty will be levied if you do.
1 List all articles you have used in your reference section (between 2 and 5 references are adequate)
3
Questions
1. Write down the research question (RQ) you will study.
For example, if your chosen aspect is board diversity in education, then the RQ will be: Does firm
financial performance depend on the diversity in educational background of the board of directors
(20 words max)
5 marks
2. White down the testable hypothesis corresponding to the RQ and the baseline regression equation
above. Include both the null and the alternative hypothesis. Carefully justify whether the alternative
hypothesis is single or double-sided. The justification must include referencesto academic literature
that develops a mechanism of how diversity affects firm performance and its direction. (200 words
max)
10 marks
3. Construct a table outlining the definitions of variables that you will use in your regression analysis.
Name this table: “Table 1: Variable Definitions”. An example table of variable definitions is given
in the Appendix. Add one more column in which you will specify the reference article(s), which
provides academic grounding for including each variable.
5 marks
4. Narrow down the data file uk_isins to include only the industries assigned to you. Merge this subset
of firms with the required datasets of financials and board characteristics.
Make sure you have examined your data for outliers and potential errors. Perform appropriate
cleaning and data management and briefly explain why and how you performed these steps. Include
in an Appendix any graphs or tables that you found helpful in this task. (200 words max)
5 marks
5. Produce tables of summary statistics of the cleaned raw variables that you will use to construct the
final variables that will be used in your regression. Name this table “Table 2: Summary statistics of
raw variables; Panel A Firm Financials; Panel B Board characteristics” and include the following
information: mean, standard deviation, minimum, median, maximum and number of non-missing
observations of each variable. (Year dummies should not be included in Table 2). Report the
number of unique firms in the notes to Panel A and unique directors in the notes to Panel B. Copy paste the lines of code in Stata you used.
5 marks
6. Based on the variable definitions question (3) construct your regression variables. Report a table
(“Table 3: Summary statistics of regression variables”) of summary statistics of the
constructed/transformed variables that will be used in your regressions. Briefly discuss whether the
number of observations and firms has shrunk relative to Table 2 and why. (100 words)
5 marks
7. Compare your Table 3 to similar tables in some of the papers from which you are drawing guidance.
Explain any notable differences. Make sure to copy/paste as an image the relevant table(s) you are
comparing your results to. (100 words max)
5 marks
8. Construct two graphs.
“Figure 1: Evolution of diversity” showing three timelines of the annual mean of your diversity
measure for your three industries on the same plot. (Hint: twoway line diversity year if ind == 1 ||
line diversity year if ind == 2 || line diversity year if ind == 3)
“Figure 2: Tobin’s q and Diversity” showing a scatterplot and a linear fit between Tobin’s q and
Diversity for all firm-year observations in your sample.
Provide a brief discussion of each figure, explaining what insights they offer. (200 words max)
10 marks
9. Construct a table displaying the mean values for two subgroups of firms: one with high and one
with low diversity and report a t-test to determine whether the means are statistically different from
each other. (Hint: construct a dummy variable high_diversity = 1 if the average diversity measure
of a firm over the period is greater than the median for the sample and 0 otherwise). Label this table
“Table 4. Differences in means. Panel A: Full sample, Panels B, C and D for each of your three
industries. Note any substantial differences between the two subgroups and among the industries,
and offer potential explanations. (200 words max)
10 marks
10. Perform regression analysis with clustering by firm. Report the following three regression
specifications in a single table (Table 5. Regression analysis). Show the Stata commands that
produce them.
Column (1) no interaction
= 0 + 0 0 + 0 +∑ +∑ +
Column (2) with interaction
= 0 + 0 0 + 0 + 1 0 +∑ +∑ +
Column (3) – same as column (1) but without industry dummies and using the FE
estimator.
10 marks
11. Provide statistical and economic interpretation of the coefficient corresponding to your testable
hypothesis. How do these interpretations change across columns (1) through (3) of Table 5
Organize your answer as in the table below corresponding to the 3 columns of Table 5 and the two
required interpretations. Indicate “same as in column (X)” if the interpretation does not change.
Make sure to provide both interpretations even if a coefficient is statistically insignificant.
Hint: A statistical interpretation states whether the null hypothesis is rejected or not and justifies
this conclusion. An economic interpretation shows the magnitude of the coefficient relative to an
informative summary statistic and makes a judgement of the relative size of the effect.
(200 words max)
4
5
Column (1)
Statistical significance Economic magnitude
Reject/Cannot reject the null hypothesis with
X% confidence because …
Firm value changes by $X mil, which is Y% of the
average firm size in Table 2 – this is tiny/huge,
realistic
Column (2)
Column (3)
10 marks
12. To answer (i), (ii) and (iii) below you will compare the regression results with Figure 2 in Question
8 and the t-test results in Question 9 in the context of your RQ. You will need to clearly specify the
coefficients being compared between the regression analysis and the differences in means analysis.
Is Figure 2 consistent with the relevant regression results and t-test Arrange your answer as
follows:
(i) Fig 2 is consistent/non-consistent with the coefficients on variables so and so in column
(X) of Table 5. Give the estimated coefficients and what aspects of Fig 2 you are comparing
them to.
(ii) Fig 2 is consistent/non-consistent with the t-test comparing the means of variables so and
so in Table 4. Give the two means and the t-test result and what aspects of Fig 2 you are
comparing them to.
(iii) Do these differences undermine or strengthen the reliability of the regression results
findings Why (200 words max)
10 marks
13. Provide plausible explanations for any of the differences highlighted in Question 12. How does
the FE estimation in column (3) help reconcile these differences Point to the exact numbers and
table you are basing your answer on. (200 words max)
10 marks
Notes:
Except for Table 1, all tables and graphs need to be produced in Stata. Hint: use the esttab command
for publishable quality tables. Images of the tables are also acceptable. Tables or graphs without clear
professional formatting and thorough informative notation will incur a penalty.
If the instructions only ask you to include a table, you do not need to provide any further discussion for
that task.
You should use Times New Roman 11 point font on A4 pages with 2.5 margins from each side. An
absolute maximum of 1500 words +/- 10% not including any tables and/or graphs, references or
appendix. Make sure to include the word count at the beginning of your write-up.
Make sure that the names you give to your variables are consistent across the report. For example, if
you refer to Log(Assets) as SIZE, then your regressions should show the coefficient for SIZE and not
for Log(Assets).
Apart from any additional graphs and tables you may have included in an Appendix, copy-paste all
commands from your do file(s) that replicate your project. The Appendix is not part of the word count.
References
Bernile, Gennaro, Vineet Bhagwat, and Scott Yonker. 2018. “Board Diversity, Firm Risk, and
Corporate Policies.” Journal of Financial Economics 127(3): 588–612.
doi:10.1016/j.jfineco.2017.12.009.
Knyazeva, Anzhela, Diana Knyazeva, and Lalitha Naveen. 2021. “Diversity on Corporate Boards.”
Annual Review of Financial Economics 13(1): 301–20. doi:10.1146/annurev-financial-
101520-065559.
Appendix
Example table of variable definitions from Bernile et al. (2018)
Note that we are asking you to have an additional column listing the article(s) that justify your choice
to include that variable and its construction.
6